Mathematical Reasoning on GSM8K (Accuracy, Token count)
96.6AccuracyVanilla
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| VanillaBackbone=Qwen3-32B2026.04 | 96.6 | 1,707 | |
| VanillaBackbone=Qwen3-14B2026.04 | 96.2 | 1,672 | |
| DTSRBackbone=Qwen3-14B2026.04 | 96.2 | 849 | |
| VanillaBackbone=Qwen3-8B2026.04 | 96 | 2,208 | |
| DTSRBackbone=Qwen3-32B2026.04 | 96 | 849 | |
| NoWAITBackbone=Qwen3-14B2026.04 | 95.8 | 1,133 | |
| NoWAITBackbone=Qwen3-32B2026.04 | 95.8 | 1,212 | |
| DEERBackbone=Qwen3-32B2026.04 | 95.5 | 979 | |
| DEERBackbone=Qwen3-14B2026.04 | 95.3 | 896 | |
| DEERBackbone=Qwen3-8B2026.04 | 95.1 | 963 | |
| DTSRBackbone=Qwen3-8B2026.04 | 95.1 | 1,152 | |
| NoThinkingBackbone=Qwen3-14B2026.04 | 95.1 | 278 | |
| NoWAITBackbone=Qwen3-8B2026.04 | 94.9 | 1,307 | |
| NoThinkingBackbone=Qwen3-32B2026.04 | 94.8 | 283 | |
| NoThinkingBackbone=Qwen3-8B2026.04 | 92.6 | 303 |